Solvient AI

Plate I — How the layer works

One base. A few agents you can name. And you, in command.

The AI operating layer for SMEs. We connect your data and tools into one base, run a few practical agents on top, and route every action through you.

Input

Your data and tools

↓

Core

The base

↓

AgentsActive

Documents●Inbox triageReporting
↓

Gate

You approve

Fig. 1 — Data enters · Base structures · Agents draft · You sign off

Book a €750 diagnosticMap your operations

Plate II — The route

A path, not a leap.

Each stage stands on its own. Start with a diagnostic; stop whenever the value is enough.

L1
AI DiagnosticFrom €750 · 3–7 days

We map where the hours go and where AI actually pays back.

L2
AI Company Base · Mini-OS2–4 weeks

Your data and tools unified into one AI-ready Mini-OS.

L3
Workflow AutomationScoped per workflow

A few named agents handle documents, inbox and reporting.

L4
Managed AI OpsMonthly retainer

We run, watch and improve the layer; you keep the controls.

L5
Multi-agent operationsOn the horizon

Agents that coordinate across the base — still gated by you.

The agents

What runs on the layer

Plate III — Documents

Drafts that read like you wrote them.

Invoices, contracts and replies, prepared from your base — never filed until you say so.

From your base

Pulls real numbers and details, not guesses.

Flags gaps

Surfaces what is missing instead of inventing it.

Your format

Matches your templates and tone every time.

Documents agentExample

“Draft the June invoice for Casa Marina from the signed quote.”

✓
Pulled rates from quote SQ-118

4 line items · €3,940 + IVA

✓
Filled client and tax details

NIF and address from the base

!
Flagged: payment terms unset

Suggesting 30 days — confirm?

● Awaiting your approvalApprove and file

Plate IV — Inbox triage

A clear inbox before your first call.

Client mail sorted, summarised and drafted — the agent proposes, you decide what goes out.

Inbox agentExample

“Triage this morning's client mail.”

✓
Booking — Hotel AlbuferaReply ready

Wants 4 PCs, Friday 18:00. Draft prepared.

✓
Billing — Gestoría VelaReply ready

Asks for May invoice copy. Attached from base.

!
Complaint — refundNeeds you

Tone-sensitive. Held for your decision.

● Awaiting your approvalApprove and send

Sorted by intent

Booking, billing, complaint — each labelled.

Drafted replies

Ready in your voice, waiting for one click.

Nothing auto-sent

Every send still passes the gate.

Plate V — Monthly reporting

The month, closed.

A reporting pack assembled from your base each month — you review before it reaches anyone.

Same day each month

Drafted on schedule, no chasing.

Traceable

Every figure links back to a source row.

Review first

Nothing leaves before you sign off.

Reporting agentExample

“Build the May management pack.”

✓
Reconciled income and expenses

142 transactions matched to the base

✓
Drafted P&L and cash summary

One page, your usual layout

!
Flagged: 3 uncategorised items

€612 total — review before sign-off

● Awaiting your approvalApprove and file

Plate VI — By hand vs on the layer

Done by hand, it is gone. On the layer, it compounds.

Manual work resets every month. A base keeps what it learns.

By hand

Redone from scratch every month

Knowledge lives in one person's head

Mistakes spotted late, if at all

Slower as the business grows

On the layer · Automated

Reused from a base that stays current

Captured once, available to every agent

Flagged at the gate, before sign-off

Faster as the base gets richer

Plate VII — Principles & safety

How we operate.

01

Operations, not code

We start from how your business runs, then bring AI to it.

02

Human in the loop

Agents prepare and propose. You approve every action.

03

A base, not a bot

One AI-ready foundation first, useful agents second.

04

Built for scale

Privacy-aware from day one, designed to travel across markets.

Safety

✓

You approve every action

Nothing leaves the building until you say so

✓

Clear data boundaries

We connect only what is agreed

✓

Privacy-aware handling

Modern data practice, residency in mind

Plate VIII — Proof

What the layer did for real businesses.

Real, operating businesses where we run the layer ourselves — not slideware. We report what we have measured, and flag what we are still validating.

A successful offline esports business

01 / Case

Gaming club · 29 PCs · audience across 5 language segments

We replaced three marketing contractors with an AI marketing department — an orchestrator plus specialist agents — and added three functions the team never had: coordination, end-to-end analytics, and trend-watching.

Up to €28K/yr
marketing spend saved
≈ €2,550 → ≈ €200 / mo · contractors → tool subscriptions
>10×
lower marketing cost
three contractors → two tool subscriptions
2.5×
lower cost per acquisition
€11.6 → €4.6 on paid channels
5 languages
of content, no new hires
one multilingual agent
Approvals2–3 days→~2 hours
Analytics1 day, by hand→~1 hour, automated

Built on data: we analysed the full customer history — 4,750 customers — and found ≈9% of them (VIPs) drove ≈65% of revenue. Budget and content refocused on keeping VIPs and finding their lookalikes, inside a closed loop — campaigns → data → analytics + trends → decision → optimisation — that keeps lowering the cost of acquisition.

An accounting back office for autónomos in Spain

02 / Case

Regulated admin service · recurring deadlines · multilingual clients

We replaced manual document prep and fragmented client communication with an AI back office — an orchestrator plus specialist agents for intake, extraction, classification, missing-document checks, deadline control, review packs and client updates — and added an audit trail, proactive gap detection and a structured human review before every filing.

€12–16K/yr
assistant cost it can avoid
vs one junior helper (€1,000–1,350/mo) — tools + reviewer instead
2.5–3×
target capacity per reviewer
~100 → 250–300+ clients · validating on first cycles
Every filing
human-reviewed
agents prepare; a person approves before submission
3 languages
client comms, no new hires
ES / EN / RU drafts under human approval
Review packsmanual assembly→automated draft + human approval
Missing docsreactive follow-up→proactive checklist

Every client period creates a structured trail — documents received, fields extracted, gaps detected, questions escalated, review pack approved and filing logged. The loop keeps cutting routine work while every critical decision and filing stays with a human reviewer. Detailed before/after numbers are being measured on the first operating cycles.

Plate IX — Start here

See where your operations stand.

A diagnostic from €750, in 3–7 days. A clear map of where AI pays back — and you keep the map whatever you decide next.

Book a €750 diagnosticMap your operations

or send us an email